cl_Matrix.hpp

Go to the documentation of this file.

/*
 * BELFEM -- The Berkeley Lab Finite Element Framework
 * Copyright (c) 2026, The Regents of the University of California,
 * through Lawrence Berkeley National Laboratory (subject to receipt of any required
 * approvals from the U.S. Dept. of Energy).  All rights reserved.
 *
 * Developers: Christian Messe, Gregory Giard
 *
 * See the top-level LICENSE file for the complete license and disclaimer.
 */
 
#ifndef BELFEM_CL_MATRIX_HPP
#define BELFEM_CL_MATRIX_HPP
 
/*
 * The contract of belfem::Matrix with either backend:
 *
 *   - Storage is column-major ( BLAS/LAPACK order ). Column j is contiguous
 *     and row i is strided, so an inner loop varies the row index.
 *   - data() returns the raw storage, which may include padding. spacing()
 *     is its inter-column stride: Blaze pads it for SIMD alignment,
 *     Armadillo does not.
 *   - Element access goes through A( i, j ). An offset computed by hand
 *     must use spacing(), never n_rows(). The length of a whole-matrix
 *     transfer ( BLAS, LAPACK, MPI ) is spacing() * n_cols(), never
 *     capacity().
 *   - data() of an empty matrix is well-defined and may be nullptr.
 */
 
#include "cl_Vector.hpp"
 
// include implementation
#ifdef BELFEM_ARMADILLO
#include "cl_AR_Matrix.hpp"
#include "op_AR_MatrixEqualEqual.hpp"
#elif BELFEM_BLAZE
#include "cl_BZ_Matrix.hpp"
#include "op_BZ_MatrixEqualEqual.hpp"
#endif
 
#include "op_MatrixPlus.hpp"
#include "op_MatrixMinus.hpp"
#include "op_MatrixTimes.hpp"
#endif //BELFEM_CL_MATRIX_HPP